Hier finden Sie wissenschaftliche Publikationen aus den Fraunhofer-Instituten.

Feature extraction as a part of automatic speech recognition a comparative study

: Jesorsky, P.

Nachrichtentechnische Zeitschrift : NTZ 29 (1976), Nr.12, S.898-901
ISSN: 0027-707X
Fraunhofer HHI ()
speech recognition; automatic speech recognition; comparative study; methods; feature extraction

This paper compares some methods of feature extraction in the field of automatic speech recognition. The efficiency of these methods is judged by the recognition rate of a classification, and the number of multiplications and divisions determines the amount of calculations. It is shown, that the best results are obtained by spectral analysis via inverse filtering, if simple linear classifiers are used and if, additionally, the amount of calculations is taken into account.